Boeing turns to AI to cull safety risks from mountains of data

Paris: Boeing is harnessing artificial intelligence to sift through mounds of data and identify potential hazards in its aircraft and airline operations, as it tries to bolster its safety culture following two fatal 737 Max crashes.

To identify and mitigate risks before they lead to accidents, injury or loss of life, the company has created a safety analytics tool that uses advanced mathematical models and machine learning, said Mike Delaney, Boeing’s chief aerospace safety officer. It’s part of a broader safety management system that Delaney and his team built in the aftermath of the tragedies.

Boeing’s reputation was badly damaged with the 737 Max accidents in 2018 and 2019. Several investigations into the crashes found the lack of a safety framework contributed to the miscommunication and other breakdowns that led to flawed designs on the aircraft. US regulators accepted Boeing’s system in December 2020, and earlier this year proposed mandating a similar approach for other aerospace manufacturers.

Data analysis is at the heart of Delaney’s effort. A detailed review of close calls on airport runways may eventually influence how Boeing designs its cockpits “- and how airlines train their pilots. Boeing released an annual update on its safety practices on Wednesday.

Boeing has started using the new data tools to study runway overruns, landing mishaps that can lead to close calls with other aircraft. It’s gained a better understanding of the safety risk “- and confidence will work on any large data set, Delaney said.

The goal isn’t just to find the proverbial needle in a haystack, Vishwa Uddanwadiker, Boeing’s vice president for aerospace safety analytics, told reporters during a briefing at the company’s Arlington, Virginia headquarters.

Boeing’s safety management system collects and monitors data from an array of internal and external sources, like design and manufacturing data, audit findings, and even reports that repair stations file to flag failed and malfunctioning parts to the Federal Aviation Administration.

Starting in March, the company began using a machine-learning algorithm that it developed jointly with the FAA to scan and mine data from the so-called “Service Difficulty Reports” for worrisome risks emerging within the global fleet. They’re written accounts of parts breakdowns that are filed by the maintenance shops, and not easily categorized.

The safety team is tracking 20 key performance measures on a weekly basis that are closely correlated to safety risks in designing, building or operation of its aircraft. For example, company officials can drill down and look for patterns and contributing factors behind “escapes,” Boeing parlance for defective parts not made to its specifications, Uddanwadiker said.